Breusch-Pagan Test

CFA Level II Quant
#1

Hi Everyone,

So I just found out recently that I passed Level I in December so I am trying to turn around and take Level II in June. I jumped right into Quant this week as I am using the Kaplan books and I am having a trouble wrapping my head around the Breusch-Pagan test (Really, I am having trouble with the whole ‘Correlation and Regression’ section). Does anyone have an easy way of explaining it, maybe something I can read or a video I would be able to watch on YouTube.

Thanks for the help!

#2

Try reading the chapters for this material in the CFA Institute books. Although I don’t particularly like the way the CFAI handles stats, it’s usually better than Kaplan. Let us know if you’re still unsure about things after reading the CFA chapters.

#3

I read this topic back in october or so but there was a free video on youtube…i think wiley had one…it helped me more than the reading and only took 2hrs lol.

#4

I highly recommend Mark Meldrum’s videos for this stuff…

The Breusch-Pagan test is to detect whether conditional heteroskedasticity exists in the regression model (meaning the variation of the residuals is dependent on the level of the independent variable. On a plot, imagine the data points being tight to the regression line for lower values but more spread out as x increases).

To test if conditional (bad) heteroskedasticity is present, you would create a new regression with the residuals from the first regression. The new Rsquared for this regression is used in the equation Rsquare * n for the BP-statistic and use K (the number of independent variables used in the regression) as the degrees of freedom to look the critical value up in the chi-square table in the book.

If the calculated BP-stat is greater than the critical value found in the table, you reject the null hypothesis and determine that there IS conditional heteroskedasticity present in the model, meaning the model may not be reliable. If you have Cond. Heteroskedasticity, you will have to take more steps to get a reliable model but that’s more than I want to type out.

#5

Although you have the right idea, remember that the assumption is about the random error term, not the residuals.